EP0137704A2 - Schwindender analoger Speicher - Google Patents

Schwindender analoger Speicher Download PDF

Info

Publication number
EP0137704A2
EP0137704A2 EP84305945A EP84305945A EP0137704A2 EP 0137704 A2 EP0137704 A2 EP 0137704A2 EP 84305945 A EP84305945 A EP 84305945A EP 84305945 A EP84305945 A EP 84305945A EP 0137704 A2 EP0137704 A2 EP 0137704A2
Authority
EP
European Patent Office
Prior art keywords
imager
hgcdte
capacitor
analog
silicon
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Withdrawn
Application number
EP84305945A
Other languages
English (en)
French (fr)
Other versions
EP0137704A3 (de
Inventor
Claude E. Tew
Cecil R. Coale, Jr.
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Texas Instruments Inc
Original Assignee
Texas Instruments Inc
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Texas Instruments Inc filed Critical Texas Instruments Inc
Publication of EP0137704A2 publication Critical patent/EP0137704A2/de
Publication of EP0137704A3 publication Critical patent/EP0137704A3/de
Withdrawn legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G11—INFORMATION STORAGE
    • G11C—STATIC STORES
    • G11C27/00—Electric analogue stores, e.g. for storing instantaneous values
    • G11C27/02—Sample-and-hold arrangements
    • G11C27/024—Sample-and-hold arrangements using a capacitive memory element
    • G—PHYSICS
    • G11—INFORMATION STORAGE
    • G11C—STATIC STORES
    • G11C27/00—Electric analogue stores, e.g. for storing instantaneous values
    • G11C27/02—Sample-and-hold arrangements
    • G11C27/024—Sample-and-hold arrangements using a capacitive memory element
    • G11C27/026—Sample-and-hold arrangements using a capacitive memory element associated with an amplifier
    • H—ELECTRICITY
    • H01—ELECTRIC ELEMENTS
    • H01L—SEMICONDUCTOR DEVICES NOT COVERED BY CLASS H10
    • H01L27/00—Devices consisting of a plurality of semiconductor or other solid-state components formed in or on a common substrate
    • H01L27/14—Devices consisting of a plurality of semiconductor or other solid-state components formed in or on a common substrate including semiconductor components sensitive to infrared radiation, light, electromagnetic radiation of shorter wavelength or corpuscular radiation and specially adapted either for the conversion of the energy of such radiation into electrical energy or for the control of electrical energy by such radiation
    • H01L27/144—Devices controlled by radiation
    • H01L27/146—Imager structures
    • H01L27/148—Charge coupled imagers
    • H01L27/14875—Infrared CCD or CID imagers
    • H01L27/14881—Infrared CCD or CID imagers of the hybrid type
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
    • H04N23/20—Cameras or camera modules comprising electronic image sensors; Control thereof for generating image signals from infrared radiation only

Definitions

  • the present invention relates to infrared imagers.
  • infrared area array imager which will provide a detailed image at wavelengths down to the limits of the available long-wavelength atmospheric window, i.e. at wavelengths of 8-12 microns.
  • Hq 1-x Cd x Te a very narrow bandgap semiconductor
  • HgCdTe a very narrow bandgap semiconductor
  • HgCdTe area arrays have tvpicallv used HgCdTe photodiodes as optical detectors, and have interconnected these photodiodes with silicon processing stages.
  • a difficulty in long-wavelength imaging is that the background flux is very high. That is, the peak black body wavelength at room temperature is very close to 12 microns, so that stray (near-field) long-wavelength radiation is likely to be generated by the infrared imaging optics of the imaging system. Moreover, verv little of a field of view will usually be trulv dark. That is, in thermal imaging the brightness variation within an image will be due to typically small variations in temperature and to variation in the black-body coefficients. These effects produce a dynamic range within a scene which is relatively small. Little of the photon flux carries information since most of the flux merely corresponds to the high average temperature of the scene.
  • the nresent invention teaches an MIS detector.
  • An infrared imager wherein the signal sensed is repeatedly averaged in a capacitor.
  • the output of the imager is thus sensed as an exponential average over a large number of read cycles, which provides a qreatly improved signal-to-noise ratio. Recursive averaging is thus optinally accomplished without a large software load.
  • the circuit which performs this analog processing is very simple.
  • the present invention will first be discussed with referencing to a whole family of embodiments using a hybrid structure, wherein the averaging capacitor sites are vertically interrated with the detector sites. Next, an additional range of embodiments, wherein the averaging sites are physically separated from the detector sites will be discussed.
  • FIGURE 6 The entire imager chip of the present invention is generally shown in FIGURE 6.
  • a HgCdTe area 106 which contains, in this embodiment, a 32 X 32 array 56 of detector sites 50.
  • each HgCdTe detector site 50 is a silicon averaging capacitor site 54.
  • the sequence of signals read from each HgCdTe detector site 50 is recursively averaged in its corresponding silicon averaging capacitor site 54, and the output from the silicon averaging capacitor is then provided to one of the output circuits shown on the left and right sides of the drawing.
  • HgCdTe detector sites 50 The fabrication and structure of the HgCdTe detector sites 50 will first be described. The preferred architecture of the silicon underneath the HgCdTe, and the preferred operation of the array, will be described subsequently.
  • FIGURE 2 A plan view of the presently preferred embodiment each of of the HgCdTe detector sites is shown in FIGURE 2. With the geometrics shown, about 35% of the total area is optically accumulative, and about 55% of the total stores charge.
  • the HgCdTe suhstrate 106 comprises lonq-wavelenqth material, e.g. HG .8 Cd .2 Te.
  • a storage gate 14 of semi-transparent metal e.g. nickel or chromium
  • the storage gate 14 is preferably thin, e.g. 100 anqstroms, to promote transparency.
  • This storage gate overlies an insulator 15, such as zinc sulphide.
  • a main via 16 connects the HgCdTe sensor portion to the corresponding silicon averaging capacitor site.
  • a via stop 18 is placed on the storage gate, and a metal film interconnect (e.q. 1 micron of indium) is used to connect the via stop 18 to the contact 22 on the silicon at the bottom of the via.
  • a monocrystalline HgCdTe slice is lapped to be parallel. At least one side is polished smooth, using one half percent bromine in methanol. The polished side is anodicallv oxidized to about 700 angstroms of oxide 58 in potasium hydroxide, using, e.g. 300 microamps current for CdHgTe slice measuring about 0.2 X 0.9 inches.
  • the HgCdTe used is n-type, with a carrier concentration of 5E14 per cubic cm. Of course, lower carrier concentrations could be used.
  • the HgCdTe is mounted on the silicon processors.
  • the HgCdTe mounting is preferably done before the silicon wafer has been completely diced up into chips, so that a HgCdTe bar can be mounted to, e.g., 4 scribed silicon chips simultaneously, before the chips are actually separated.
  • the silicon substrates (chips or groups of chips) are preferably flat, and must be cleaned thoroughly on both sides. Each silicon substrate is heated to a controlled temperature of 100C, and a very low-viscosity and low-surface tension glue 60 is then applied to the center detector pattern (i.e. to the averaging capacitor area on each of the silicon chips) on the silicon bar.
  • the HgCdTe bar is then placed on the silicon substrate, oxide down.
  • the alignment marks in the silicon are used to obtain rough alignment of the HgCdTe bar at this time.
  • a press is used to hold the HgCdTe in place while the glue sets, using, e.g., three to four ounces of pressure over the 37 X 32 array.
  • the gluing step is required to produce a thin and uniform glue bond, and a low viscosity epoxy glue such as Epotek 301 is preferably used. After the gluing operation, the excess HgCdTe which hangs over the edges of the silicon bar is then scribed and broken off.
  • the silicon substrate, having the HqCdTe bar glued to it, is now waxed to a glass lapping plug.
  • These plugs are commercially available parts, formed, e.g., as cylinders one inch in diameter and 1 3/4 inches tall, having flat and parallel ends.
  • the flatness of the substrate on the lapping plug is preferably within two microns. If it is not flat within two microns, the waxing step is preferably performed over. The thinning steps can then begin.
  • the alignment marks on the silicon are protected, e.g. by painting them with resist and curing, prior to the fitting operation.
  • the remainder of the silicon processor bar should be adequatelv protected by its thick (preferably one micron) nitride protective overcoat.
  • HgCdTe It may also be desirable to protect the contact pads on the silicon bar during the polishing step.
  • exposed resist is applied over the whole structure prior to polishing.
  • the resist on the HgCdTe will be polished off.
  • the HgCdTe is then lapped, e.g. with three micron grit on glass down to about 75 microns thick, followed by lapping with 0.3 micron grit on a polishing cloth (e.g., Buehler "Texmet" TM) until the scratches left from the three micron lap are gone.
  • the HgCdTe is then polished, with half percent bromine-methanol on a pellon pad, down to 20 to 25 microns.
  • a final polishing step, using one eighth percent bromine-methanol solution takes the HgCdTe down to I? microns above the silicon surface.
  • the silicon substrate is then wax-mounted to a silicon secondary substrate, which is used simply as a carrier for processing.
  • a silicon secondary substrate which is used simply as a carrier for processing.
  • sapphire or other rigid material can be used for this carrier.
  • the structure in hand is a long thinned strip of HgCdTe, glued across several connected Si chips.
  • the first patterning step is now performed, to reduce the area 106 required for the imaging array 56 on each chip.
  • a small strip of HgCdTe is also preferably left in place connecting each of the arrays together. This provides electrical connection between the HgCdTe areas, and is useful to pass current during anodic oxidation. It will be removed later during processing.
  • this step is done, e.g. by spinning on a resist such as 180CP (available from Hunt) at 4,000 RPM for 20 seconds.
  • the resist is dried for three minutes at 60 C, exposed, developed for thirty seconds, rinsed for 15 seconds, and ashed at 200 watts for 30 seconds.
  • special protection to any exposed aluminum, such as the alignment marks, since aluminum is attacked by bromine-methanol.
  • a spray etch with fresh 1/8% bromine-methanol is then applied until the pattern clears, i.e. until the separation trenches have etched down to the epoxy.
  • the remaining thin epoxy is then ashed off on a plasma asher, and the photoresist is then removed.
  • the second patterning step patterns the via 16 through the HgCdTe. This is a crucial step, since the via 16 geometry must satisfy several constraints. If the via 16 is etched too small, the device can short out at the bottom of the via. If the via is etched too large, so that it intrudes into the active area of the device, the dark current will be qreatly increased. Third, the walls of the via 16 must be very smooth to provide good step coverage. For example, simple ion milling leaves rough walls. Thus, the presently preferred embodiment uses ion milling to make a small via hole (e.g. 0.4 mils), then uses spray etching to enlarge the via to get to the right diameter.
  • ion milling to make a small via hole (e.g. 0.4 mils)
  • spray etching to enlarge the via to get to the right diameter.
  • the ion milling is preferably done with argon ions, using e.g., 600 milliamps beam current at 600 volts at about 1/4 micron of pressure.
  • the stage is preferably angled ten degrees to the beam and is rotated. This stage must be cooled. In the presently preferred embodiment the stage is cooled to about 5 C during milling.
  • the ion milling is performed intermittently, to assure that the device does not heat up locally. Heating is most undesirable because the dis-sociation of HgCdTe occurs at relatively low temperatures. Such dissociation can alter the bandgap, degrade interface stability, increase the carrier concentration, or have other undesirable effects.
  • the device After ion milling has cut all the way through the HgCd T e, the device, with its photoresist still intact, is spray etched in fresh 1/8% bromine-methanol in ten second stages until the via has been undercut to a total diameter of (in the presently preferred embodiment) 0.7 mils. After the photoresist is removed, the whole device is then static etched in one-eighth percent bromine methanol for fifteen seconds, to remove any damaged layer which may have been generated during the ion milling step.
  • the next step is growth of an anodic oxide layer 62 on the HgCdTe.
  • oxide is grown to first blue (around 700 angstroms thickness) in potassium hydroxide, using a current of, e.g., 23 microamps for a 1-device substrate. This provides a mixture of cadmium oxides, Te oxides, and tellurites.
  • This layer plus the zinc sulfide layer 15 form the insulator of the MIS capacitor.
  • Resist is now patterned to define the areas where the first zinc sulfide layer 15 is to be deposited. The zinc sulfide is deposited everywhere except on the silicon.
  • the zinc sulfide lifted off, e.g by immersion and spray in I tetrachloroethylene.
  • ZnS over anodic oxide is the preferred insulator, but many others (preferably of high permittivity) can be used, e.g.
  • the transparent MIS gate 14 can now be deposited.
  • a thin i resist layer (e.g. 4000 or 5000 angstroms) is formed over areas outside of the storage gate 14.
  • a short ion milling step is applied for adhesion before the gate is actually deposited.
  • One hundred twenty five angstroms of nickel is then deposited, and lifted off from undesired areas.
  • the via stop 18 is then formed, again by lift off.
  • Preferably 800 angstroms of aluminum are used.
  • an adhesion milling step is preferably used before the metal is deposited.
  • a second zinc sulfide layer 64 is then deposited in the same locations as the first zinc sulfide layer. Preferably 9000 angstroms are deposited in the second layer. Preferably in adhesion milling step is used before the zinc sulfide is deposited. Again, the zinc sulfide is patterned by liftoff.
  • the MIS via 66 is cut. This will cut the opening for connection of the metal interconnect 20 to the via stop 18 on the storage gate 14. This is preferably done using a 1/8% bromine-methanol wet etch.
  • the substrate via is preferably cut.
  • the substrate via is not shown in FIGURE 2, but provides the bias connection for the HgCdTe barlog, and thus for the MIS capacitor. (preferably the HgCdTe bar is biased at about -5V with respect to the silicon substrate.) This etch is preferably done using 1/8% bromine methanol.
  • the main via 16 is then opened again, using the same mask as in the original via etch step (the second mask).
  • a 1/8% bromine-methanol etch is preferably used. This etch is used to open up contact to the aluminum bond pads 22 in the underlying silicon, as shown in FIGURE 1. Note that, at the bottom of via 16, a substantial width of the first insulator layer 15, 14 and the second insulator layer 64 separates the H g CdTe substrate 106 from the interconnect 20 and band pad 22. This is due to the overetch applied during initial cutting of the via 16.
  • an alternative embodiment of the present invention uses p-type rather than n-type HgCdTe material, again at a carrier concentration of, e.g., 5E14 per cubic centimeter (although the actual carrier concentration may, in p-type as in n-type embodiments, be anywhere from 5E16 down to 1R13 or less).
  • a carrier concentration e.g., 5E14 per cubic centimeter
  • positive trapped charge in the oxide does not have a channel stop effect, but has precisely the opposite effect. That is, the space charge of electrons whidh is found all over the surface of the HgCdTe 106 will immediately flow into the depl.etion wells underneath the storage gates 14, swamping the desired signal provided by photoelectrons.
  • one additional masking step is used to deposit a field plate over the whole surface of the HgCdTe except for the storage gates 14 are found.
  • the mask used for the field plate is preferably the complement of the mask used for the MIS gates 14, or is slightly larger. That is, a small degree of overlap can be provided between the field plate and the storage gates 14.
  • field plate is not restricted to embodiments using p-type material, but can also be used in many other embodiments, and, in particular, desirable in any embodiment where the insulating laver does not contain a sufficient amount of trapped charge of the proper sign.
  • Figure 15 shows an embodiment using a field plate 70.
  • FIG. 4 A further embodiment of the present invention, in which bump bonding rather than via connections is used to connect the narrow-bandgap semiconductor to the silicon substrate, is shown in Figures 4 and 5.
  • this embodiment there is still exactly one averaging capacitor sight located in the silicon underneath each detector sight in the HgCdTe, but bump bonding is used rather than via methods to form a connection. That is, as shown in Figure 5, each storage gate 14 would be connected through a bonding strap 108 to a tall indium post 104.
  • the respective indium posts 104 are then pressed onto a pattern of the contact pads 22 on the silicon, and, preferably with slight heat applied to assist fusion, the indium posts 104 are cold-welded onto the contact pads in the substrate.
  • a low viscocity epoxv glue is preferably used to fill the void in the remaining spaces above the substrate.
  • thinning of the HgCdTe is removed from the constraint imposed by via geometry.
  • the diffusion length of carriers in HgCdTe is normally extremely long, collection is still most efficient when the location where photogeneration of the carriers occurs is not too far removed from the location where collection occurs.
  • this structure can be fabricated epitaxially. That is, using a cadmium telluride (CdTe) substrate, a thin epitaxial layer of HgCdTe is used for the active device structure. Where bump bonding is used to connect such a structure, the CdTe substrate can be as thick as is desirable for mechanical handling properties, since CdTe is transparent to wavelengths from about 1 micron to about 30 microns.
  • CdTe cadmium telluride
  • a photodiode structure rather than an MIS capacitor can be used as the detector sight. That is, a device structure such as shown in Figure 16 would be used at the HgCdTe level.
  • the material used is preferably p-type material having a carrier concentration of, e.g., lEl5 per cubic centimeter.
  • n-type layer 114 which is preferably less than 1 micron thick (again, it is desirable to have the sight of photogeneratin of carriers be close to or coincident with the sight of their collection. Even long-wavelength photons are normally absorbed in a semiconductor very close to the surface, e.g. within 1 ⁇ 2 micron.)
  • This structure can be formed, for example, after the HgCdTe material has been bonded to the silicon and thinned as described above, by cutting the via hole 16 down to the aluminum contact patterns 22 and silicon s described before. At this point processing diverges.
  • n-type layer 114 To form the n-type layer 114, almost any species of ion can be implanted, since the implant damage, rather than the particular species implanted, tends strongly to produce n-type material in HgCdTe.
  • a species which is not a dangerous impurity in silicon is preferably implanted, and the species implanted need not even be a dopant impurity in HgCdTe.
  • argon ions can be implanted to produce n-type layer 114.
  • the walls of the via hole 16 are also implanted, so that formation of an ohmic contact from the n-type layer 114 to the contact pad 22 on the silicon can be very easily performed, simply by depositing a small amount of metal 118 (such as indium over nickel) at the bottom of the via hole 16. (Of course, it is preferable to use metals which are not as dangerous to silicon as nickel)
  • metal 118 such as indium over nickel
  • the same mask which is used to cut the via hole 16 can be used to pattern the metal 118. Since it is very easy to form ohmic contacts to HgCdTe, and particularlv n-type HgCdTe, the choice of the metal 118 is very easy.
  • the HgCdTe is etched away to form mesas 120, as shown in Figure 17. Again, it is necessary to have some ground connection for the p-type HgCdTe substrate.
  • An alternative way of connecting the n-tvpe regions 114 to the aluminum pad 22 on the silicon is through a via which is not located within the active device mesa 120, but is adjacent to it, as shown by the vias 122 in Figure 17.
  • the photodetector site represents a depletion capacitance, instead of the series combination of a depletion capacitance with an MIS gate insulator.
  • the qeneral mode of operation of the photodiode embodiment is fairly similar to that of the MIS embodiment. That is, a bias voltage is applied to the photodiode which opens up a wide depletion region at the junction boundary. As p hotocarriers are generated, this depletion region will be narrowed, thus modifying the capacitance. If the capacitance is sampled after a fixed time interval, its value will indicate the photon flux received.
  • a bias voltage V b is first aoplied to both the photodiode and the averaging capacitor C h . After the photodiode has had time to collect flux, it is connected to C b for integration, and then isolated and biased to begin another collection period.
  • this mode of operation is substantially different from the mode of operation used in the prior art of photodiode device. That is, the prior art applies a constant bias to a photodiode, and often includes much complicated circuitry to achieve the constant bias. In effect, the prior art senses changes in current caused bv changes in flux, and the bias voltage is therefore carefully selected to achieve a locally maximum value of the derivative of current with respect to flux.
  • the key criterion for sensitivity is the partial derivative of capacitance with respect to flux, across the band of voltages to which the photodiode is exposed during its integration time. That is, as flux generates photocarriers in the photodiode, the width of the depletion region will gradually narrow, and therefore the capacitance across the diode will increase and the voltage will decrease.
  • the maximum flux collection time in the photodiode is limited not merely by the considerations of number of integrate cycles per read cycle and the desired timing of the read cycles, as discussed above with reference to the principle preferred embodiment, but is also limited by the self-debiasing behavior of the photodiode as discussed above.
  • the metal interconnects 20 are patterned. Preferably these two are patterned by liftoff. Preferably a double adhesion mill is performed and 100 angstroms of nickel is deposited first, and then about 11 ⁇ 2 microns of indium are deposited (using angle sputtering). At this stage also are formed the ground connection for the HqCdTe bar 106, through the substrate via.
  • the bond pad area 22 within the metallization layer 68 is defined by the absence of the protective nitride overcoat, which otherwise protects all of the silicon chip.
  • Bond pad 22 is aligned with via 16 in the CdHgTe cell overhead, so that the metal interconnect 20 connects bond pad 22 to storage gate 14.
  • the metal bond pad 22 is connected to a moat 36 at contact 42, and transistors 38 and 40 provide selective connection to the bond pad.
  • the bond pad 22 is connected to the bottom plate 32 of the poly-to-poly capacitor 30.
  • the poly-to-poly capacitor 30 has a capacitance which is typically at least comparable to storage gate 14, and thus, when transistor 38 is activated, the signal which has been generated between storage gate 14 and the CdHgTe substrate can be averaged with the signal which has been stored in capacitor 30, as will be discussed below.
  • the transistor 40 is controlled by another clock line p hi mux , and this transistor selectively connects the bond pad 22 (and therefore the storage gate 14) to a read line 24.
  • the read line 24 is used to apply these two signals to all cells.
  • both lines p hii and phi mux are raised, to connect bottom plate 32 to word line 24.
  • the fabrication of the silicon is preferably done using a 3 micron double-poly MOS process, which is standard except in one point: the dielectric of capacitor 30 is preferably formed as an oxide/nitride/oxide composite.
  • This particular composite dielectric provides a precisely control.led high specific capacitance together with a low defect density.
  • an oxide is formed over first poly (either grown or deposited), nitride is partially oxidized at the same time as the second gate oxide is grown. This process is described in greater detail in U.S. Patent Application 468,Q20 (TI-9335), which is hereby incorporated by reference.
  • FIGURE 12 shows a hypothetical schematic which will be used to discuss the effect of recursive averaging.
  • the switches S1 and S2 are opened alternately, with no overlap.
  • the output voltage will be referred to as a time series X n
  • the input voltage will be referred to as a time series Y n .
  • switch S 2 When switch S 2 is closed, the output voltage X n equals a weighted average of the previous output voltage with the current input voltage: where alpha is a factor which is defined as: ⁇
  • X n is equal to Y n :
  • the signal-to-noise ratio is then rhus, recursive filtering enhances the signal-to-noise ratio by an amount which depends on the factor alpha, i.e. on the ratio of the input capacitor C a to the averaging capacitor C b .
  • C a in this example corresponds generally to the MIS capacitor defined by the gate 14, and C b corresponds generally to the averaging capacitor 30 which is located in the silicon. That is, the recursive averaging can substantially reduce the noise power which is due to the background flux.
  • FIGURE 9 shows the layout and timing of a first simplified embodiment of the present invention. Note that the vias shown at every pixel location connect vertically to the HgCdTe MIS detector site which is not shown.
  • This simplified embodiment depicts only a 3 X 3 array, but much larger array sizes are preferable, and array sizes of 128 X 128 and larger are expected to be the best mode of practicing the present invention in the future.
  • the Ml. switches are used both to reset the MIS array and to multiplex the signals during readout. The M2 switches perform the recursive integration onto the silicon averaging capacitors.
  • the phii ( p hii n t e g ra t e ) signals are operated in unison for all pixels in order to simplifv clocking. Separate voltages are supplied for resetting the read line ( Vr ) and for emptying the MIS wells (Ve).
  • Vr read line
  • Ve emptying the MIS wells
  • a typical clocking sequence for such an array is as follows: (1) V e is impressed onto all MIS capacitors via transistor ME and all transistors Ml. This collapses all wells, and the storage charge is recombined. (2) V r is impressed onto all MIS capacitors via transistor MR and all transistors Ml. This allows optically generated charge to collect in all of the MIS wells.
  • the well of a row being read can be collapsed. That is, if the wells are not collapsed before readout, the incident flux will be changing the voltage across the MIS capacitor, and therefore the output voltage, while the read operation is in progress. It is conceivably possible that, under some flux conditions, such a changing output signal might trigger an instability in some sense amplifier design. However, this mode of operation is believed not to be a necessary precaution in the preferred embodiment.
  • the preferred additional circuitry uses a correlated double sampler. This correlated double sampler is clamped just after the reset voltage has been applied, and is released before the row line is activated. After the read output has been provided to the read line, the second sample is taken. The signal information is found by comparison of the empty well potential (which is equal to V r ) and the actual potential which is right out from a particular storage gate 14.
  • the timing chart of FIGURE 13 shows the timing of the read operation in slightly more detail.
  • the duty cycle and sensitivity are not optimal, but explanation of this embodiment will clarify the operation of the preferred recursive read mode of operation which will be described next.
  • the signal which appears on any one row line phi n is the logical sum of a common row line signal phi com with a component pheta n which is peculiar to that row line.
  • event number 1 which is the start of the nth cycle, all detector elements are first reset to V e , which puts them in accumulation.
  • the theta n signal selects the nth row. All detectors are connected to their respective integration capacitors 30 by a phii signal. After the rise of both theta n (and therefore phi n ) and phii, the nth detector's capacitor 14 and the nth averaging capacitor 30 (C b ) are both reset to V r . The reset voltage V r is chosen so that the storage capacitor 14 will be depleted. During event 3, photogenerated carriers are collected in the nth row detector. At event 4, the collected charge in the nth row detector is sensed. The output signal is the difference between the detector voltage at the end of event 2, which is the empty-well potential V r , and the detector voltage at event 4.
  • the averaging capacitor 30 has a capacitance which is at least twice that of the MIS capacitor. This translates to a value of 2/3 for the parameter alpha, and a maximum signal-to-noise improvement of the square root of 5.
  • the number of rows which needs to be integrated during each active integrating time is at least one plus alpha over 1 minus alpha, or in this case five rows. After a row has been read, it can be left idle until it can usefully begin collect-and-integrate cycles preparatory to the next read phase.
  • detector N remains addressed through signal theta N , while rows N+l through N+4 are isolated from the read line.
  • signal phi i connects all capacitors 30 to their respective detectors 14. Those that are connected to active detectors again recursively integrate the latest photo generated signal with the stored averaged signal.
  • the output signal is again defined by the difference between full and empty well potentials.
  • a five frame recursive average of the nth row is found on the read line just prior to the phi r clock at event 4. Several more cycles of timing are illustrated to show the standby integrate and readout progression throughout the array.
  • Figure 11 shows a sample preamplifier circuit, used to amplify the signal from one detector pixel.
  • M 9 is used to isolate preamp from detector array during Det. probe.
  • M 1 and M 2 are source followers.
  • M2 and M4 are enhancement inverting gain stages. Capacitor C bandlimits the system.
  • M 5 and M 6 are output buffer source followers.
  • V b biases M 3 & M 6 as current sources.
  • the present invention can tolerate low quality HgCdTe device material, since it is less sensitive to the noise figure degradation of dark currents than most prior art devices are.
  • the present invention can be embodied using HgCdTe having intrinsic carrier concentrations of 5X10 14 or higher, although this is not preferred.
  • HgCdTe device Quality is such that a 1500 angstroms zinc sulphide MIS gate can sustain a voltage differential of (V-V th ) of 1 volt, and a 3000 angstrom layer of zinc sulphide can withstand 2 volts. Assume dark current to be 300 microamps per square centimeter for the storage area.
  • HgCdTe having a intrinsic carrier concentration in the neighborhood of 5E14.
  • the 8-10 micron spectral band is received.
  • the processor input preferred noise specification is set at 10 microvolts, and that the load capacitance added bv the silicon input circuit is 2 picofarads.
  • the net transmission through the optics is 0.7, and that the maximum flux in the focal plane (assuming a 350 0 K farfield seen through f/3.2 optics having 8" focal length) is 1.6 E6 photons per centimeter squared per second.
  • the dielectric constant of zinc sulphide is 8.8.
  • the load capacitance is 2.3 picofarads at the pixel.
  • the pixel integration time must be 12 microseconds.
  • the fullwell charge capacity is 3.3E6 carriers.
  • the duty cycle is 9.2%, assuming currently used frame rates typical of a 64x64 infrared area imager.
  • the sensitivity is then 538 microvolts per degree centigrade at the pixel site, and 237 microvolts per degree C at the output node.
  • the noise at the read node is calculated as 14 microvolts, and the noise equivalent temperature figure for 1 frame is 0.059 0 C.
  • the system equivalent noise temperature at a 60 Hz framerate is then 0.0170C.
  • the maximum data rate on one silicon column is equal to the length of the array, divided by the number of recursive integration times plus 1, divided by the integration time. If 9 recursive integrations are used, the integration time is 12 microseconds, and each column contains 64 pixels, then the maximum data rate is seen to be 530 kHz.
  • Maximum system sensitivity can be acquired by summing frames for a relatively long period of time, e.g. 16 msecs. That is, one frame can be read in (N+!)(ti+t a ), where N is the number of recursive integration cycles, and ti and t a are the integration and addressing required times respectively. For example, if 9 recursive integrations are applied, and 12 microsecond integration is possible, and 4 microseconds are required for addressing, then one frame can be read in 160 microseconds. Therefore, in 16 msecs 100 frames can be read, and the system equivalent noise temperature is further .improved by a factor of the square root of 100. This results in an estimated system equivalent noise temperature of 0.0059°C.
  • the present invention advantageously provides: an infrared area imager which will provide a detailed image at wavelengths down to the limits of the available atmospheric window, i.e. at wavelengths as long as about 12 microns.
  • This invention performs the required integration recursively (exponential integration) in the analog domain. In nd add doing such, the necessity of high speed analog to digital conversion is eliminated. In fact, if the integrated analog information needs to be digitized for another system related reason, the data rate can be reduced according to the degree of integration performed in the analog domain.
  • Figure 12 shows a simplified schematic diagram of the recursive action of the memory cell.
  • the continuous voltage and continuous time signal is sampled and held on Ca by Sl. when S2 closes, the charge stored on Ca and Cb is shared.
  • a solution to insure alpha being as designed is to make alpha externally adjustable.
  • Such a scheme is shown in Figure 26. It should be clear that a Cb exists for each pixel; however, since the video is usually serial in time from pixel to pixel, Ca can be time division multiplexed to the appropriate Cb, i.e., one Ca could run the whole NxM memory or, say, N Cas could run the NxM memory. In either event, Ca is repeated much fewer times than Cb. Therefore, it is feasible to make Ca adjustable by closing Sx and/or Sy and/or Sz. In doing such, the value of Ca can be trimmed to suit the requirement. This proposition holds true if one desires to make alpha adjustable for system related requirements.
  • Fig. 19 shows a proposed architecture for implementing this invention.
  • Two circuits as shown perform frame integration on the output from a 64 by 64 focal plane imager.
  • the operating temperature is 77 0 K .
  • This benefits operation in that leakage currents drop to near zero and need not be considered during the design.
  • Room temperature (300K) or thermo-cooled (200K) operation are alternative embodiments, but increased leakage currents will result.
  • This analog memory can easily be constructed using MOSLSI technology.
  • the switches are simple MOS transistors while the capacitors are realized using two level polysilicon parallel plates. This method of construction uses significantly less gate area than a comparable CCD structure. Unlike a CCD analog memory, it also is highly forgiving of processing tolerances.
  • the processor was designed to operate with the 64x64 HgCdTe CID described in U.S. Patent 4,360,732, which is hereby incorporated by reference. However, this processor can also be used in conjunction with many other kinds of infrared imagers.
  • the architecture consists of 32 channels of preamplification followed by individual correlated double samplers (CDS) and synchronous clamps. The outputs of the CDS and synchronous clamps each have buffers consisting of a source follower configuration.
  • the present invention advantageously provides a long-wavelength detector which provides a very good signal-to-noise ratio, even in the face of substantial. long-wavelength background flux.
  • the preamp ( Figure 20) used was of a cascode configuration.
  • the design parameters are given in Table 1 below. All devices in Figure 20 are n-channel enhancement mode transistors.
  • the gain of the cascode is approximately equal to the square root of the W/L ratios of the driver (Ml) and the load (M3). Since the voltage gain of Ml (from gate to drain) is about -2.3, the Miller multiplier is held to 3.3. This fact, along with the self aligned gate structure of Ml, allowed low input capacitance to be achieved, for a standard 800 A NMOS process.
  • M4 and M5 form an output buffer (A v approximately .9) for the gain stage.
  • M6 is used to reset the read line. Cinj is the inject capacitor.
  • the bandwidth of the amplifier is determined by the g m of the load transistor M3 and the rolloff capacitor, C rolloff ⁇
  • the correlated double sampler is used to eliminate KTC (Reset) noise and output offsets.
  • Figure 21 is a schematic of the CDS.
  • C clam p is the clamp capacitor and C s/h is the sample capacitor.
  • Ml is the clamp switch, M2 is the sample switch, and M3 is the switch which sets the bias for the output buffer, consisting of M4 and M5.
  • the tradeoffs here are: 1) the attenuation of the signal through the voltage divider, C clamp and Cs/h; 2) C s/h has to be sized such that KTC noise is negligible compared to the preamp output noise.
  • the synchronous clamp is shown in Figure 22. Its purpose is to remove column correlated offsets and to set the bias of its output buffer, comprised of transistors M2 and M3.
  • the synchronous clamp operation degrades the SNR by a factor of root (2); i.e., the output noise out of the CDS is increased by 1.4. This extra noise can be virtually eliminated by bandlimiting the synchronous clamp.
  • the analog memory uses switched capacitors to perform a first order recursive filter action on video samples.
  • the ratio C b/ C a +C b determines the integration feedback coefficient, alpha.
  • Figure ?3 shows one column of eight For an individual preamp channel.
  • R 1 through R 8 as well as C 1 through C 8 are generated by on-chio memory control logic.
  • the memory has a common input/output line buss for each channel.
  • the data stored in memory is an analog voltage in discrete time.
  • the first order recursive filter improves SNR by a factor of , where a is the integration feedback coefficient.
  • the noise gain throuqh the memory is while signal gain is 1.
  • the memory output buffer serves two functions.
  • the S/H circuit serves as a temporary analog memory that stores data; thus the output data rate can be reduced.
  • C d is used to take out effects due to V t variations in Ml, shown in Figure 24, and also shifts the quiescent output of M6.
  • M 6 is the driver for the output multiplexer.
  • the multiplexer shown in Figure 25, is a common load type multiplexer, where M5 is the common load for all 32 channels.
  • Video data, out of the memory output buffer, are commutated onto the gate of the output driver M10, by switches Ml, M2, M3, ..., M32.
  • the switches are controlled by a two-phase ratioless shift register. Phase 3 is used to reset the analog switches and to establish a reference for the output buffer.
  • the Fading Memory Processor was manufactured using a modified NMOS process which consisted of a 600 angstrom gate oxide and a 400 angstrom interlevel oxide (poly 1 - poly 2). It was necessary to use a 400 angstrom interlevel to achieve a suitable alpha (integration feedback coefficient) for a SNR improvement, in the analog memory, of approximately 2.25.
  • the 600 angstrom gate oxide was used so that 1/f noise in the cascode preamp driver could be reduced, where C ox is the gate oxide capacitance and t ox is the thickness of the gate oxide.
  • the results of 77°K testing of one slice are shoen below in Table 2.
  • the power dissipation was excessive, being approximately 65 mW.
  • VDD and VHI were reduced to +8 V to bring the power dissipation under the spec and goal parameters of 50 mW.
  • the only loss in doing this was a decrease in input dynamic range (200 mV to 150 mV).
  • Preamp Dynamic Range was designed for approximately 250 mV (input referred). Because of the decrease in input dynamic range due to power dissipation requirements and the slightly larger preamp gain (29 as opposed to 25), the input dynamic range was decreased to 150 mV. However, since the input referred noise was 6.5 uV, the preamp dynamic range approached the goal of 90 dB.
  • the present invention advantageously provides a long-wavelength detector which provides a very good signal-to-noise, even in the face of substantial long-wavelength background flux.
  • the present invention provides a major novelty in the art of infrared imagers.
  • the present invention can be embodied in a wide variety of modifications and variations, and the present invention is not limited except as may be defined in the claims. All claims should be interpreted broadly, in accordance with the pioneering status of the present invention.

Landscapes

  • Engineering & Computer Science (AREA)
  • Power Engineering (AREA)
  • Physics & Mathematics (AREA)
  • Computer Hardware Design (AREA)
  • General Physics & Mathematics (AREA)
  • Condensed Matter Physics & Semiconductors (AREA)
  • Electromagnetism (AREA)
  • Microelectronics & Electronic Packaging (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Solid State Image Pick-Up Elements (AREA)
  • Transforming Light Signals Into Electric Signals (AREA)
  • Closed-Circuit Television Systems (AREA)
  • Photometry And Measurement Of Optical Pulse Characteristics (AREA)
EP84305945A 1983-08-31 1984-08-30 Schwindender analoger Speicher Withdrawn EP0137704A3 (de)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US52830483A 1983-08-31 1983-08-31
US528304 1990-05-24

Publications (2)

Publication Number Publication Date
EP0137704A2 true EP0137704A2 (de) 1985-04-17
EP0137704A3 EP0137704A3 (de) 1988-07-20

Family

ID=24105111

Family Applications (1)

Application Number Title Priority Date Filing Date
EP84305945A Withdrawn EP0137704A3 (de) 1983-08-31 1984-08-30 Schwindender analoger Speicher

Country Status (3)

Country Link
EP (1) EP0137704A3 (de)
JP (1) JPS60155932A (de)
IL (1) IL72696A (de)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
FR2599529A1 (fr) * 1986-06-03 1987-12-04 Telecommunications Sa Procede d'integration dans un circuit integre monolithique des courants de sortie de diodes de detection et integrateur pour la mise en oeuvre du procede
EP0507541A2 (de) * 1991-04-04 1992-10-07 Texas Instruments Incorporated Verarbeitungsschaltung mit Integrator und Tiefpassfilter für jeden Pixel einer infraroten Bildebene-Matrix
EP0660600A1 (de) * 1993-12-23 1995-06-28 Thomson-Csf Spannungsaddierer und Addierer-Mosaik für ein Wärmebildgerät
EP0801399A2 (de) * 1996-04-08 1997-10-15 Sony Corporation Analoge Verzögerungsschaltung
EP0813339A2 (de) * 1996-06-11 1997-12-17 Canon Kabushiki Kaisha Festkörperbildaufnahmevorrichtung
EP1139563A2 (de) * 1996-05-28 2001-10-04 Matsushita Electric Industrial Co., Ltd. Analoger FIFO-Speicher
KR100822097B1 (ko) * 2004-03-18 2008-04-15 가시오게산키 가부시키가이샤 비디오 재생 장치, 재생 방법 및 비디오 재생 프로그램을 기록한 기록 매체

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4786809A (en) * 1986-07-31 1988-11-22 Santa Barbara Reseach Center Low noise preamplification stage
JP4781154B2 (ja) * 2006-04-14 2011-09-28 中国電力株式会社 脱塩装置からイオン交換樹脂を取り出す方法及び装置

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0019269A1 (de) * 1979-05-17 1980-11-26 Licentia Patent-Verwaltungs-GmbH Verfahren und Schaltungsanordnung zur Trennung des thermischen Hintergrundsignals eines IR-Detektors vom Nutzsignal
WO1982001789A1 (en) * 1980-11-10 1982-05-27 Barbara Res Center Santa Charge coupled device open circuit image detector
US4379232A (en) * 1977-12-19 1983-04-05 Texas Instruments Incorporated Ferroelectric imaging system
EP0137988A2 (de) * 1983-08-31 1985-04-24 Texas Instruments Incorporated Infrarot-Bildwandler

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4379232A (en) * 1977-12-19 1983-04-05 Texas Instruments Incorporated Ferroelectric imaging system
EP0019269A1 (de) * 1979-05-17 1980-11-26 Licentia Patent-Verwaltungs-GmbH Verfahren und Schaltungsanordnung zur Trennung des thermischen Hintergrundsignals eines IR-Detektors vom Nutzsignal
WO1982001789A1 (en) * 1980-11-10 1982-05-27 Barbara Res Center Santa Charge coupled device open circuit image detector
EP0137988A2 (de) * 1983-08-31 1985-04-24 Texas Instruments Incorporated Infrarot-Bildwandler

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
IEEE TRANSACTIONS ON ELECTRON DEVICES, vol. ED-29, no. 1, January 1982, pages 3-13, IEEE, New York, US; K. CHOW et al.: "Hybrid infrared focal-plane arrays" *

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
FR2599529A1 (fr) * 1986-06-03 1987-12-04 Telecommunications Sa Procede d'integration dans un circuit integre monolithique des courants de sortie de diodes de detection et integrateur pour la mise en oeuvre du procede
EP0507541A2 (de) * 1991-04-04 1992-10-07 Texas Instruments Incorporated Verarbeitungsschaltung mit Integrator und Tiefpassfilter für jeden Pixel einer infraroten Bildebene-Matrix
EP0507541A3 (de) * 1991-04-04 1992-11-25 Texas Instruments Incorporated Verarbeitungsschaltung mit Integrator und Tiefpassfilter für jeden Pixel einer infraroten Bildebene-Matrix
EP0660600A1 (de) * 1993-12-23 1995-06-28 Thomson-Csf Spannungsaddierer und Addierer-Mosaik für ein Wärmebildgerät
FR2714501A1 (fr) * 1993-12-23 1995-06-30 Thomson Csf Sommateur de tensions, et mosaïque de sommateurs, pour appareil d'imagerie thermique.
EP0801399A2 (de) * 1996-04-08 1997-10-15 Sony Corporation Analoge Verzögerungsschaltung
EP0801399A3 (de) * 1996-04-08 1999-04-21 Sony Corporation Analoge Verzögerungsschaltung
EP1139563A2 (de) * 1996-05-28 2001-10-04 Matsushita Electric Industrial Co., Ltd. Analoger FIFO-Speicher
EP1139563A3 (de) * 1996-05-28 2002-01-02 Matsushita Electric Industrial Co., Ltd. Analoger FIFO-Speicher
EP0813339A2 (de) * 1996-06-11 1997-12-17 Canon Kabushiki Kaisha Festkörperbildaufnahmevorrichtung
US6130712A (en) * 1996-06-11 2000-10-10 Canon Kabushiki Kaisha Eliminating the influence of random noise produced by an optical black pixel on a reference output
EP0813339A3 (de) * 1996-06-11 1998-10-14 Canon Kabushiki Kaisha Festkörperbildaufnahmevorrichtung
KR100822097B1 (ko) * 2004-03-18 2008-04-15 가시오게산키 가부시키가이샤 비디오 재생 장치, 재생 방법 및 비디오 재생 프로그램을 기록한 기록 매체

Also Published As

Publication number Publication date
IL72696A (en) 1987-12-20
JPS60155932A (ja) 1985-08-16
EP0137704A3 (de) 1988-07-20

Similar Documents

Publication Publication Date Title
US4684812A (en) Switching circuit for a detector array
US4779004A (en) Infrared imager
US4686373A (en) Infrared imager
US4862237A (en) Solid state image sensor
US6930336B1 (en) Vertical-color-filter detector group with trench isolation
US9142585B2 (en) Image sensor comprising isolated germanium photodetectors integrated with a silicon substrate and silicon circuitry
US6943838B2 (en) Active pixel sensor pixel having a photodetector whose output is coupled to an output transistor gate
US7687302B2 (en) Frame shutter pixel with an isolated storage node
US6476374B1 (en) Room temperature, low-light-level visible imager
EP0137704A2 (de) Schwindender analoger Speicher
US5235197A (en) High photosensitivity and high speed wide dynamic range ccd image sensor
EP0137988B1 (de) Infrarot-Bildwandler
Tsaur et al. 128x128-element IrSi Schottky-barrier focal plane arrays for long-wavelength infrared imaging
US5270558A (en) Integrated electronic shutter for charge-coupled devices
JPH0778959A (ja) 固体撮像素子
US8105858B2 (en) CMOS imager having a nitride dielectric
WO2006043105A1 (en) Electro-optical device
Broudy et al. Infrared/Charge Coupled Devices (IR/CCD) Hybrid Focal Planes
Kozlowski et al. Visible and infrared detectors at Rockwell Science Center
JPH0456272A (ja) 固体撮像装置
Reich et al. High performance charge-coupled-device imager technology for plasma diagnostics
Vural et al. Large-format SWIR/MWIR HgCdTe infrared focal plane arrays for astronomy
Haas et al. 2048x2048 Visible/Infrared HgCdTe Focal Plane Arrays (FPAs)
McNutt Monolithic Schottky-barrier focal plane arrays with high fill factors
Meisenzahl et al. 31 Mp and 39 Mp full-frame CCD image sensors with improved charge capacity and angle response

Legal Events

Date Code Title Description
PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

AK Designated contracting states

Designated state(s): DE FR GB

PUAL Search report despatched

Free format text: ORIGINAL CODE: 0009013

AK Designated contracting states

Kind code of ref document: A3

Designated state(s): DE FR GB

17P Request for examination filed

Effective date: 19890109

17Q First examination report despatched

Effective date: 19900411

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE APPLICATION IS DEEMED TO BE WITHDRAWN

18D Application deemed to be withdrawn

Effective date: 19900822

RIN1 Information on inventor provided before grant (corrected)

Inventor name: TEW, CLAUDE E.

Inventor name: COALE, CECIL R., JR.